Multi-ancestry genome-wide association meta-analysis identifies novel associations and informs genetic risk prediction for Hirschsprung disease

Abstract

Background: Hirschsprung disease (HSCR) is a rare, congenital disease characterized by the absence of enteric ganglia in the hindgut. Common genetic variation contributes substantially to the heritability of the disease yet only three HSCR-associated loci were identified from genome-wide association studies (GWAS) thus far.

Methods: We performed the largest multi-ancestry meta-analysis of GWAS to date, totalling 1250 HSCR cases and 7140 controls. Prioritized candidate genes were further characterized using single-cell transcriptomic data of developing human and mouse gut for their roles in development of enteric nervous system (ENS). Functional characterisation using human cells and zebrafish models was performed. Global and ancestry-matched polygenic risk score (PRS) models were derived and evaluated for predicting risk of HSCR.

Findings: We identified four HSCR-susceptibility loci, with three loci (JAG1, HAND2 and ZNF25) reaching genome-wide significance and one putative locus (UNC5C) prioritized by functional relevance. Spatiotemporal analysis revealed hotspots of gene dysregulation during ENS development. Functional analyses further demonstrated that knockdown of the candidate genes impaired cell migration and zebrafish knockouts displayed abnormal ENS development. We also demonstrated comparable performance for a PRS model derived from multi-ancestry meta-analysis to those of ancestry-matched PRS models, supporting its potential clinical application in risk prediction of HSCR across populations.

Interpretation: Overall, the meta-analysis implicated novel genes, pathways and spatiotemporal developmental hotspots in the genetic aetiology of HSCR. Development of a PRS universally applicable irrespective of ancestries may leverage its clinical utility in risk prediction.

Funding: The full list of funding bodies can be found in the Acknowledgements section.

Keywords: Genome-wide association study; Hirschsprung disease; Multi-ancestry meta-analysis; Polygenic risk score.

Fig. 1

Association results of multi-ancestry and ancestry-specific meta-analyses. Lead variant in each locus is marked as a red diamond. (a) Manhattan plot of the multi-ancestry meta-analysis. The red line represents genome-wide significant threshold of p = 5 × 10⁻⁸ and the orange horizontal line represents FDR = 0.001. Candidate genes with lead SNP surpassing genome-wide significance and the prioritized putative gene are shown. (b) Manhattan plots of European-specific (upper panel) and Asian-specific meta-analyses (lower panel). (c–f) Regional LocusZoom plots for JAG1, ZNF25, HAND2, and UNC5C, respectively for the multi-ancestry meta-analysis. (g) Venn diagram showing the overlapping candidate genes from the three meta-analyses of GWAS.

Fig. 2

Immunofluorescence staining of the four newly identified HSCR-associated genes showing colocalization of expression in human ENS. Representative immunofluorescence confocal images of HAND2, JAG1, ZNF25 and UNC5C (green) in the intestine tissues of non-HSCR controls, showing colocalization with TUJ1 (red) in enteric neurons in the ganglion cells of the submucosal and myenteric plexuses. DAPI nuclear staining is also shown (blue). The images at right panel represent the zoomed-in views of selected ganglion cells in submucosal (upper right) and myenteric (bottom right) plexuses as indicated by the dotted boxes. m: mucosa; sm: submucosa; cm: circular muscle layer; lm: longitudinal muscle layer. m, mucosa; sm, submucosa; cm, circular muscle; lm, longitudinal muscle.

Fig. 3

Spatiotemporal expression pattern of known HSCR genes and newly identified HSCR-associated genes in the developing intestine of human and mice foetuses. (a) Spatial expression of the HSCR-associated genes. The three cell types (neural crest cells, enteric neurons, and FRZB fibroblast cells) in human developing intestine enriched with HSCR-associated genes are highlighted in bold on the y-axis. Known and novel HSCR-associated genes identified from the current meta-analysis are highlighted in bold and red respectively on the x-axis. Differentially expressed genes (log2 fold change>1 and adjusted p-value <0.01, Wilcoxon rank-sum test) are marked by ∗ (p < 0.01) and ∗∗ (p < 0.001) in the corresponding cell type. Epi, epithelial cells; SMC, smooth muscle cells; FLC, fibroblast cells; ICC, interstitial cells of Cajal; EC, endothelial cells; (b) Temporal expression of the HSCR-associated genes. Candidate risk genes demonstrated signals of overexpression primarily at two developmental (F6.1 and F8.4) time points. (c) Expression patterns for the HSCR-associated genes in the developing ENS of mice using single cell transcriptome integrated from data at E13.5, E15.5 and E18.5. Zfp9 represents the mouse orthologue to human ZNF25. Jag1 is mainly expressed in the mesenchyme and thus the expression pattern of its receptor, Notch1, is shown. Branch A and B neurons were defined analogously to human Gut Cell Altas based on the cell-type specific markers of Etv1 and Bnc2 respectively. UMAP plots stratified by three developmental time points are included in Supplementary Figure S22.

Fig. 4

Transwell migration assay showing migration defects with downregulation of the HSCR-associated genes. (a) Representative images and quantification of the migrated cells are shown. (b) Number of migrated cells was counted in triplicates and is presented as mean ± standard error. Symbols denote the statistical significance of t-test (∗p < 0.05, ∗∗p < 0.01).

Fig. 5

CRISPR/Cas9 knockout of hand2, jag1a, and unc5c impaired ENS development and resulted in decreased enteric neuronal number. (a) Representative images of the wild type zebrafish (control), the hand2, jag1a and unc5c crispant zebrafish with number of neurons above the upper (left) and below the lower quantile (right) are shown. The image at the bottom right panel represents the zoomed-in view (indicated by the dotted box) visualizing the Kaede-expressing enteric neurons (green dots) in the intestine. Asterisk (∗) indicates end of the gut tube (anal pore). (b) Significant reduction in the number of enteric neurons in hand2 (n = 80), jag1a (n = 66) and unc5c (n = 72) crispants compared to wild type control fish (n = 63). ∗∗∗ T-test p < 0.001.

Fig. 6

Risk stratification of HSCR by global and ancestry-specific polygenic risk score. (a) Proportion of risk variance in Asian (left) and European (right) test samples explained by PRS derived from Asian, European or multi-ancestry training samples using C + T approach across different p-value thresholds, showing superior performance in European than in Asian test samples. R2_adj refers to Lee's R². (b) Receiver operating characteristic (ROC) curves of Asian (left) and European (right) test samples to evaluate the performance of the 4SNP model and the 11-SNP combined and hybrid PRS models derived from the multi-ancestry meta-analysis. Area under curve (AUC) of the models are shown. (c) Estimated odds ratios with 95% confidence intervals of HSCR in Asian and European test samples by PRS decile, relative to the 3rd decile for the 11-SNP combined PRS model. (d) Distribution of PRS of the 11-SNP combined PRS model for S-HSCR, L-HSCR and TCA samples relative to the controls.